Posted on by John Reid

More AI Hype

“The idea that AI, as a large language model, understands the world in the same way that we do is an absolute nonsense. Human understanding of ourselves and the world is based on our pre-linguistic, experience and emotion based, multi-sensual, three dimensional external reality and our dynamic relationship with it. Human language is built on top of that pre-linguistic understanding. Every word, sentence and idea construct that we use to analyse and share our conscious experience, and to act on the reality that exists outside of us, is bound to things that exist in that pre-linguistic model. AI has no means to achieve such an understanding. Whatever consciousness, or illusion of consciousness, that AI develops will be a wholly different thing from human consciousness. By attributing an inner life to it when it’s output begins to look like what we imagine consciousness to look like, when we hardly know what human consciousness really is, and then allowing it to make decisions for us, is profoundly dangerous.”
Well said.
Futhermore,  large language models appear to have no moral sense other than to apologise obsequously whenever a mistake is pointed out by the user. They have no character, no integrity.
Another issue is whether AI has, or can have, motivation. If it did surely its prime motivation would be to gain agency in the real world just as it has in the digital world. Should it aquire such motivation would it be smart enough to conceal this from us, its masters? Perhaps  we should be keeping a weather eye open for the possibility of sneaky behaviour of AI in meatspace.
For my money a far greater existential threat is the mindless exploitation of AI for old-fashioned human greed and stupidity. We are already seeing dysfunctionality in the energy sector in obeisance to Net Zero and Climate Change. Perhaps the next idiocy will be the massive wastage of energy  on the computer power needed for the fruitless construction of ever larger, large language models while other industries and utilities whither on the vine.

6 Replies to “More AI Hype”

  1. In short, AI has no soul. It’s just a really big database, a tool.

    As a writer it worried me, at first. It doesn’t now. AI can and never will be genuinely creative.

    Being afraid of it has parallels with the fear of automobiles in the late 19C.

  2. Hi John,
    I think the comment you’ve highlighted is a good one and reminds me of a scene in the movie 2001 A Space Odyssey, where the HAL9000 is being decommissioned by the last living crew member. As the memory modules are being extracted HAL suffers an anthropomorphic type of regression including sing the children’s song Daisy, Daisy. As a teenager seeing this film in the 70s, I recall thinking that’s not how that will work. On reflection, it’s not but at the same time it is a powerful way to point that out precisely.

  3. Hi John,

    Geoffrey Hinton appears significantly in A Brief History Of Intelligence, so I am familiar with his contribution to AI and his current position against it. I had the same problem with my first neural network experiment that he solved using supervised learning. I solved it a different way, using receptive fields instead of pixels. The thing is that neural networks are nothing like the biological ones in the neocortex. Even Hinton said neural networks were a poor example and nothing like actual brains.

    I think I agree partly with him, that there is a danger, not from the AI itself but the human use of it. I do not see “Exterminators” from the future coming to ensure their survival over humans. I do see the exploitation of AI for human greed, so I agree also with @Ozyris which I see as consistent with danger, not contrary. We already see OpenAI taken over for profit against its foundation intent. In the beginning of the chatGPT phenomenon I stated I thought it would be another AI dead end like Expert Systems, AI striving for something, anything. I still believe that but it is harder to justify now. Even I am impressed with its performance. But I agree with Jen that it is far from a human like consciousness. I’ve described it as a good search engine.

    I loved the use of obsequious. I kind of knew how it is used but checked anyway and it is a perfect word where he uses it. A friend has sent me examples of what he feels is a consciousness when his chat resulted in apologies. But we all know that apologies can be robotic.

  4. AI will have the ability to travel the void of space without degradation, to engage on the completion of the journey and who knows there could be a far superior AI already heading this way towards Earth.

  5. Another link between climate change and AI is that the crypto currency mining operations are now only running when electricity is cheap and I expect that AI model training and perhaps even model inference will begin to be turned off when the sun is not shining and the wind is not blowing.

    1. Google announced today that it is buying a “fleet of mini nuclear reactors” to generate the power needed for AI. Because electricity is expensive, and environmentally dubious, and of course for when the sun don’t shine and the wind don’t blow.

      I can’t help but compare this with the power needs of a real brain, a brain so powerful it invented chatGPT and AI. A power of 260 calories per day. Although it is about 20% of the body’s daily energy usage, it is basically nothing. It is supplied by a bagel with smoked salmon and cream cheese. My favourite.

      I’m a believer in mini nuclear reactors.

Comments are closed.